Does likelihood function of your choice impact the asymptotic posterior of MCMC?

12 Views Asked by At

The metropolis Hasting algorithm decides whether to jump based on posterior probability. Namely, likelihood x prior.

Then it seems the density function that you choose (e.g., Poisson or Gaussian) can have quite some impact on the asymptotic posterior distribution we obtain from MCMC, because I can imagine there might be a case where for likelihood function L1, the parameter set $\theta_1$ has higher likelihood than parameter set $\theta_2$, whereas for the other likelihood function L2, the relation is reversed (i.e., $L1(\theta_1) > L1(\theta_2)$, but $L2(\theta_1) < L2(\theta_2)$.

Then $\theta_1$ would gain more density in the asymptotic posterior with L1, whereas $\theta_2$ would gain more density in the asymptotic posterior with L2. Different choice of likelihood function might give different asymptotic posterior. Am I right with this?